AI and advisory
I advise infrastructure and asset-intensive clients on digital, data and AI work, and my group builds and tests the methods. Client names and project data stay with the clients.
Method
Standards first
AI must not invent data, confidence is stated, people check every output, client data stays out of public services, and AI-written code is reviewed before use. Wider use comes second.
Division of work
AI produces first drafts and repeated checks; the engineer reviews, decides and signs. In my group I specify the requirements, review the code and accept the results; implementation is by the team and AI coding agents.
Audit trail
On safety-critical assets, a model is worth as much as the evidence that it can be trusted. We build that in: independent re-computation, signed audit ledgers, benchmark audits, and written records of tests that have not yet passed. Contractual targets are kept apart from achieved results.
Start small
Scope starts from the decision the work must support. Acceptance, data ownership and whether any AI tool may see client data are written down before work begins. I would then test the approach on one work package before wider use, and hand over methods and standards so the client can continue alone.
Record
Client names withheld. Roles as written in the contracts.
Automated compliance checking, 2024
Won in open competition, with me as project lead: rule-base research for automated compliance checking of building models, for a large infrastructure operator. Accepted after the client's expert review in December 2024.
Ageing assets, 2025–2027
Principal investigator on three successive commissioned research contracts on how ageing steel lattice structures fail under extreme wind and how to assess, strengthen and repair them. I lead a multidisciplinary team, including full professors, building digital-twin prototypes and AI surrogate models.
Digital transformation advisory, 2023–2024
Co-lead of the consultancy team, in the bid and in delivery, for a six-month advisory on digital transformation and BIM-based digital delivery, with an external standards expert as lead.
Information standards to repeat business, 2022–2025
For a commercial developer I led a BIM implementation guideline on information standards, quality checklists and data sharing, then, as principal investigator, a tendered façade quality assessment cross-checked by two independent algorithms, delivered and accepted. In 2025–26 the developer group commissioned further work through the testing organisation it had appointed.
Prototypes
Research prototypes and demonstrations, not systems in client use.
Assurance prototype, 2026
Re-computes analysis results through an independent chain and keeps a signed audit ledger. It has caught a real error in the main chain: a load counted twice in a load combination.
Working demonstration
FacilityNexus, on the Lab page: a 3-D digital-twin view of a building with an operations assistant, running on open datasets (CC BY 4.0) for invited testers.
Published
Corresponding author of seven journal papers published in 2026; these are the closest to AI practice.
Buildings, 2026
A language model drafts candidate rules from a knowledge graph; low-confidence rules go to expert review. Corresponding author. doi:10.3390/buildings16040719
Applied Sciences, 2026
An audit of a widely used benchmark, with a duplicate-aware evaluation split. Final and corresponding author. doi:10.3390/app162010430
Buildings, 2026
Corrosion grading from images, linked to reliability analysis of ageing steel lattice structures. Corresponding author. doi:10.3390/buildings16112250
Applied Sciences, 2026
Explicit gates for source time, spatial extent and solver balance, so each input can be traced. Corresponding author. doi:10.3390/app16199420
Applied Sciences, 2026
Multi-scale wind coupled with several forms of degradation. Corresponding author. doi:10.3390/app16136788
Buildings, 2024
A modified detection network for helmet wearing on construction sites. First author. doi:10.3390/buildings14123883
Working with me
AI assurance
Where AI can be trusted in an asset decision and what evidence it needs, including dataset and benchmark integrity audits and independent verification of AI-assisted engineering results.
Data
A data maturity assessment, a digital capability review and a data governance framework: who owns what is produced, how confidential material is handled, and what AI may see.
Teams
Written rules first and wider use second, so the team's practice improves without risking client data.
Capability building
As Founding Dean of the Smart City and Sustainable Development Academy (SCSDA), I built an applied research and advisory institute from nothing, with 12 research centres and laboratories built with partners, four overseas, and signed its client contracts as legal representative.